An invasion revisited: the African big-headed ant (Pheidole megacephala) in northern Australia
- First Online:
- Cite this article as:
- Hoffmann, B.D. & Parr, C.L. Biol Invasions (2008) 10: 1171. doi:10.1007/s10530-007-9194-x
- 341 Views
Long-term studies provide the best information for invasion ecology as multi-temporal sampling can illuminate invasion dynamics, such as population changes and rate of spread. In 1996, Hoffmann et al. surveyed an infestation of African big-headed ant, Pheidole megacephala, in semi-natural rainforest of northern Australia. Here, we re-survey this infestation 9 years later to assess the dynamics of this invasion against the key finding of the initial study. Importantly, we re-sample a site sampled prior to invasion to demonstrate the causal link between an infestation and changes in a native invertebrate community. We found the area infested had almost doubled, and P. megacephala biomass in infested sites was up to 18 times greater than that of native ants in uninfested sites. In the two sites with the youngest infestations in 1996, P. megacephala abundance had increased more than 20-fold than that measured in 1996. Native ant abundance and species richness in infested sites were notably lower in 2005, with only one native ant specimen found in the most recently infested site. The abundance of other macro-invertebrates was the lowest in the three oldest infested sites. Coleoptera and Orthoptera were less abundant in infested sites. This study supports the findings of the first ‘snapshot’ study, and has shown that the area infested has almost doubled, populations are increasing, and ecological impacts remain severe. It reinforces that P. megacephala represents a serious ecological threat, and we argue that there is a greater need for management of this ant globally.
Invasion research typically involves impact studies with ‘snapshot’ comparisons of invaded and uninvaded areas (Moller 1996), lacking pre-invasion data of the affected communities. In such cases, it is not clear if differences in native diversity between invaded and uninvaded sites are the cause or consequence of invasion (King and Tschinkel 2006). Although short-term studies are valuable, it is clear that the spatial dynamics and ecological effects of invasions can vary greatly through time (Morrison 2002), which cannot be detected or demonstrated by one-off sampling. Long-term studies provide the best information, particularly when they include explicit comparisons of the same areas before and after invasion. Repeated sampling also has the added advantage of being able to illuminate temporal aspects of invasion dynamics, especially changes in population densities and rate of spread.
Ants are recognised as globally significant exotic invaders (Williams 1994), and over the past few decades research on them has driven much of the understanding into biological impacts on invaded communities and invasion mechanisms (Moller 1996). However, much of the multi-temporal work has been conducted within agro-ecosystems, focusing on ant community mosaics and the interaction of exotic ants with phytophagous fauna, especially relating to their potential as biological control agents (e.g., Greenslade 1971; Majer 1976; Beardsley et al. 1982). Except for recent work on the Argentine ant Linepithema humile (e.g., Holway 1998; Suarez et al. 1998; Wetterer et al. 1998; Krushelnycky et al. 2005), few studies have described ant invasion dynamics within natural or semi-natural environments (e.g., Erickson 1971; Haskins and Haskins 1988; Morrison 2002; Abbott 2005).
In 1996, Hoffmann et al. (1999) conducted a detailed survey of an infestation of the African big-headed ant, P. megacephala, in a rainforest patch near Darwin in Australia’s Northern Territory. Pheidole megacephala is listed as one of the word’s worst invasive species (Baskin 2002) because of its ability to displace native invertebrate faunas (Haskins and Haskins 1965; Lieberburg et al. 1975; Majer 1985; Heterick 1997), to adversely affect agricultural production (Bach 1991; Cudjoe et al. 1993; Jahn and Beardsley. 1994), and because of its capacity to be a serious domestic nuisance (Jarvis 1931; Brimblecombe 1958; Hoffmann 2003, 2004). Like most invasion studies, the Hoffmann et al. (1999) survey suffered from a lack of pre-invasion data, with differences in invaded and uninvaded sites inferred as being due to P. megacephala. The study found (1) the ant infested 25 ha, with the distribution centred on the drainage lines and structurally complex rainforest, (2) P. megacephala abundance in infested sites was 37–110 times that of total native ant abundance in uninfested sites, and (3) the abundance and richness of native ants and other invertebrates were significantly reduced where P. megacephala was present, inversely relative to the abundance of P. megacephala.
Here we report results from a re-survey of this infestation 9 years later (including a site that was not previously infested), in order to demonstrate the causal link between an infestation and changes in a native invertebrate community, and to provide insights into the dynamics of this invasion over time. Specifically, we investigate change in (1) the distribution and rate of spread of this invasion, (2) the biomass of P. megacephala at infested sites, and (3) the impacts of P. megacephala on native invertebrates, especially ants.
The study was conducted throughout Howard Springs Nature Park and Hunting Reserve, approximately 35 km east of Darwin in Australia’s Northern Territory (12°27′ S, 130°58′ E). The Nature Park covers an area of 283 ha and surrounds a spring that feeds a permanent creek surrounded by monsoon rainforest. The Hunting Reserve extends from the northern end of the Nature Park, covering an area of 1,605 ha (CCNT 1990).
Rainfall in the region is highly seasonal, with 90% of the approximately 1,500 mm annual average falling between November and March (Taylor and Tulloch 1985). Mean daily temperatures vary from 25 to 33°C during the wet season and 17 to 30°C during the dry season. Mean relative humidity ranges from 72–83% in the wet season and 32–62% in the dry season (Bureau of Meteorology, Darwin).
Evergreen monsoon rainforest, surrounding a permanent creek, dominated by the trees Polyalthia australis, Morinda citrifolia, Gmelia schlechteri and Terminalia microcarpa, and the palms Carpentaria acuminata and Livistona benthami to about 20 m height.
Semi-evergreen Dry Vine-thicket, associated with seasonally dry substrates bordering the rainforest at its southern end. Dominant trees were Acacia auriculiformis, T. ferdinandiana and Planchonia careya in the over-story (up to 10 m in height) and Exocarpus latifolius, Brachychiton megaflora, and Flemingia parviflora in the mid-story.
Open Shrubland, occurring on poorly drained, coarse silaceous sand, bordering the rainforest in all other areas except the southern end. Dominant tree species included Eucalyptus papuana, Pandanus spiralis, Xanthostemon paradoxus, Melaleuca spp. and Grevillea pteridifolia, generally below 5 m in height, with annual Sarga spp. dominating the under-story.
The recreation area contains maintained lawns, rangers’ residences and other buildings, but most of the remaining area remains intact despite some minor use prior to the establishment of the Reserve in the 1980s and the presence of some weeds, including Stachytarpheta sp., Mission Grass (Pennisetum polystachion) and Hyptis (Hyptis sauveolens). It is not known exactly how or when P. megacephala was introduced in the Park, but assuming that the infestation was initiated around the recreation area, and using the rate of spread information determined by Hoffmann et al. (1999) with its known distribution along the creek, it is believed to have been in the Park at least 33 years.
Pheidole megacephala distribution
Pheidole megacephala is a particularly good study organism for invasion research, as like most other tramp ant species it is unicolonial, whereby dispersal involves the migration of a newly inseminated queen and a small group of workers to generally within a few feet of the parent colony (Hölldobler and Wilson 1990; Passera 1994). On a local scale, this dispersal method limits its range expansion, and provides the opportunity for detailed measurements of rate of spread.
The extent of the distribution of P. megacephala was determined using similar methods to those used in Hoffmann et al. (1999). The presence of P. megacephala (i.e., individual foragers and nests) was visually inspected along 28 transects spaced approximately 100 m apart, radiating out from the permanent creek until the ant could no longer be found (Fig. 1). The exact limit of the ant’s distribution was confirmed by attracting the ants to tuna baits (teaspoon size) placed every 2 m for a further 20 m from where the ants were last observed. The baits were inspected after approximately half an hour for the presence/absence of P. megacephala. Only one location of the limit of the ant’s distribution in 1996 could be confidently re-determined due to landmarks in each of the three habitat types. All survey work was conducted during the cooler parts of the day (600–1000 h and 1600–1900 h), when temperatures (24–28°C) do not hinder ant activity.
Ecological impacts and biomass
Pheidole megacephala impacts on native macro-invertebrates (defined as arthropods >1 mm) were quantified in the rainforest only using pitfall traps and using the methodology of Hoffmann et al. (1999). Four grids of traps were established in areas infested with P. megacephala, and another four in uninfested areas (control sites). Three of the infested sites were infested sites in the 1996 survey (sites I1, I2, I3; Hoffmann et al. 1999), and one was an uninfested site in the 1996 survey (site I4; site UI1 in Hoffmann et al. 1999). Sites I1–I4 follow an age gradient of oldest to youngest infested, respectively, and were separated by 815, 330 and 15 m, respectively (Fig. 1; see also Hoffmann et al.1999). All uninfested sites sampled here (U1-4) were different to those sampled in 1996 as the prior sites were mostly located in other discrete rainforest patches. These new sites were situated 168–700 m progressively further from the invasion front along the creek and will allow for future long-term studies of this invasion.
Pitfall traps were plastic containers (internal diameter of 42 mm), filled three quarters with 70% ethyl glycol as a preservative. Twenty traps were established at each site, arranged in a grid of two parallel transects of 10 traps, with 2 m spacing between traps and approximately 20 m between transects. Traps were set on July 25th 2005, and operated for 48 h.
Due to the poor state of Australian ant taxonomy, most species could not be named. Undescribed species were assigned number codes previously published for the region (e.g., Andersen 1991), as was the case in Hoffmann et al. (1999). A full collection of voucher specimens is held at the CSIRO Tropical Ecosystems Research Centre in Darwin. Other macro-invertebrates were sorted to ordinal level only.
Ant biomass was calculated by first measuring the dry weight of a subset of individuals of each species following oven drying at 60°C for 48 h, and then multiplying the weight to account for the total number sampled.
The non-parametric analysis, Mann–Whitney U-test was used to compare ant abundance, species richness and biomass, and abundance of other macro-invertebrates at ordinal level (at site level) between infested and uninfested areas, and between uninfested sites in 1996 and 2005. This test was used because the predominantly dichotomous nature of the data between infested and uninfested areas (complete presence or absence of either P. megacephala or native ants) and limited number of sites resulted in data that did not satisfy homogeneity tests.
While there is an increased probability for type 1 error due to multiple comparisons being performed, the Bonferroni correction is notoriously conservative, and the specific correction is dependent upon the number of categories accepted for analysis which can differ within the same experiment depending on the selection criteria (Wright 1992). From an inference perspective, it can be argued that each ordinal group analysed here can be considered independent, in which case separate inferences for each ordinal group should be used (Hochberg and Tamhane 1987). For this reason, as well as the statistical likelihood of only a single spurious result, we considered that the use of the lower probability level of significance of 0.025 sufficiently alleviated this issue.
Data for ants and other macro-invertebrates were combined in a multivariate ordination to explore differences in composition of the invertebrate assemblages between infested and uninfested sites using Primer (Clarke and Gorley 2001). A similarity matrix of sites was constructed from the abundance data for all ant species (excluding P. megacephala) and macro-invertebrate orders using a Bray–Curtis Association Matrix. Sites were then ordinated using non-metric multidimensional scaling. Analysis of similarity (ANOSIM) was used to test for clustering of sites according to infested and uninfested. ANOSIM uses non-parametric permutation procedures applied to (Bray–Curtis) similarity matrices based on rank similarities between samples. ANOSIM returns an R-statistic which gives a measure of how spatially distinct groups are, with values ranging from −1 to 1, most commonly 0 to 1. The closer the R-value is to 1 the more separated the groups are in ordination space, while a value close to zero indicates no spatial separation of groups (Clarke and Warwick 2001).
Distribution and rate of spread
Pheidole megacephala was found to occupy 45 ha (Fig. 1), an increase of 80% (20 ha) since 1996. The invasion remained predominantly concentrated along drainage lines and as such maintained its general linear shape. Range expansion from the three known points of the 1996 boundary was 120 m in both the dry vine thicket and open shrubland (mean of 13.3 m per year) and 200 m in the rainforest (mean of 22.2 m per year).
Biomass and ecological impacts
Mann–Whitney U-tests of ant and other invertebrate data between infested and uninfested sites
Abundance of Pheidole megacephala
Abundance of native ants
Species richness of native ants
Other macro-invertebrates (abundance)
All macro-invertebrates (total)
Other macro-invertebrates (ordinal richness)
Other macro-invertebrate (relative abundance)
Invertebrates other than ants
This study provides the first multi-year analysis of distribution and ecological effects for any invasive ant species on mainland Australia. These multiple samples have provided a clearer insight into the dynamics of an invasion with time, and in particular have demonstrated clear changes in invertebrate communities known prior to and after invasion. The latest results provide important comparisons with the three key findings of the 1996 data.
Distribution and rate of spread
Within 9 years, P. megacephala almost doubled the area it infests from 25 to 45 ha. This increase was predominantly along the drainage lines within rainforest, which provided a shaded, moist habitat. The spread within the rainforest was nearly double that found in the drier shrubland and dry vine thicket, consistent with findings in 1996. Manipulative studies of the abiotic constraints of P. megacephala are yet to be performed, but the patterns shown here, and in other locations where it co-occurs with L. humile (e.g., Lieberburg et al. 1975; Haskins and Haskins 1988) indicates similar moisture constraints to those known for L. humile (Way et al.1997; Wetterer et al. 1998, 2006; Krushelnycky et al. 2005; Menke and Holway 2006).
Population levels of invasive ants in new landscapes commonly exceed those of the native community, but may or may not be accompanied by an increase in ant biomass. For example, Porter and Savignano (1990) found Solenopsis invicta had populations 10- to 30-fold greater than that of native ants in uninvaded plots in central Texas, with a clear (albeit unmeasured) increase in ant biomass. Alternatively, Holway (1998) found a 4–10 times increase in Argentine ant abundance compared to native ant abundance levels, but total biomass did not alter due to the small size of L. humile relative to the larger ants they were displacing.
We found P. megacephala abundance to be 11–62 times greater than that of all native ants in uninfested sites, corresponding to a 4- to 18-fold increase in ant biomass. This biomass increase is substantially lower than the abundance increase due to the small size of P. megacephala relative to that of the native ants. However, because the arboreal ant fauna in the rainforest consists predominantly of relatively larger bodied ants (Oecophylla, Calomyrmex, and Polyrhachis) than those on the ground layer (Pheidole, Cardiocondyla, and Paratrechina) (Hoffmann et al. 1999), and because P. megacephala has large impacts on the arboreal ant fauna (Hoffmann et al. 1999), it is very likely that there would be a substantial decrease in ant biomass throughout the canopy in invaded areas accompanying this increase in ant biomass on the ground layer.
The ability of invading species to populate to biomass levels far in excess of those of native species is largely unexplained, but is hypothesised to be due to a mixture of greater foraging and feeding efficiencies coupled with a break in competitive trade-offs, as well as removal from ecological constraints (Moller 1996; Holway 1999). In particular, associations with phytophagous insects and exploitation of extrafloral nectaries have been linked to inflated ant abundances and positive feedback loops, especially for invasive ants. The collapse of the rainforest ecosystem on Christmas island due to the association between the invasive Yellow crazy ant Anoplolepis gracilipes and the Red lac scale Tachardina aurantiaca (O’Dowd et al. 2003) is probably the most dramatic and the most well-known example. Less well known but similar mutualisms between P. megacephala and the soft scale Pulvinaria urbicola are also believed to be causing the complete collapse of Pisonia grandis forests on coral atolls within Australia’s Great Barrier Reef (Hoffmann et al. 2004).
At Howard Springs, P. megacephala can be seen tending coccids on sub-surface roots of the Carpentaria palm Carpentaria acuminata and leaf axils of many trees (Hoffmann, personal observation), but the populations of these phytophagous insects do not appear to be inflated to support such a high abundance of ants, although there may be greater populations both in the canopy and under the ground that have gone unnoticed.
Many long-term studies in invasion ecology have also found an initial surge in population of an invading species can be followed by a decline to equilibriated levels (Simberloff and Gibbons 2004). For example, Majer and de Kock (1992), studying the rehabilitation of sand mines in South Africa, found that after 5 years P. megacephala abundance reached a peak of 97% of all ant abundance, but then decreased until year 13 when native ant abundance and community composition was more characteristic of undisturbed forest.
The populations of P. megacephala are clearly not declining or stabilising at Howard Springs, rather they have increased with those at sites I2 and I3 being 24 and 29 times greater, respectively, in 2005 compared to 1996. The patterns of P. megacephala abundance also differ greatly between the two sample times. The population in 1996 followed a distinct declining trend with decreasing age of infestation (distance along the rainforest away from site I1) with an extremely elevated population at the oldest infested site I1, which provided strong negative correlations of impacts on all invertebrates (Hoffmann et al. 1999). In 2005, we found it had a more uniform abundance throughout the length of the rainforest, with a shift in the point of greatest abundance from I1 to I2. But the pattern of impacts, particularly within multivariate analysis, clearly has an “ecological footprint” that shows a trend of time since invasion (i.e., I1–I4).
Invading species are often implicated with homogenizing native faunas (Clavero and Garcia-Berthou 2006; Holway and Suarez 2006), but the degree of impact on the ant community found here is amongst the most severe reported from anywhere, being an almost complete extirpation of the epigaeic fauna rather than a homogenization. Only one native ant individual was found in the youngest infested site, and no other native ants were found in any of the older sites. This is in stark contrast to the 1996 samples where many more native species were found in the youngest invaded sites. While a comparison cannot be made for the other non-ant macro-invertebrates (as data were not recorded from pitfall traps in 1996), the effects of invasion clearly trend with time since invasion (I1–I4) as was found by litter samples and foliage beats in 1996.
The underlying mechanisms by which P. megacephala is able to exert such competitive asymmetry on ants in invaded communities remain to be studied in detail, like has been done for L. humile (Holway 1999; Thomas and Holway 2005; Menke and Holway 2006). Nevertheless it is very likely that P. megacephala has greater exploitative abilities than most natives by having quicker recruitment times, recruiting in higher numbers, foraging for longer periods, having higher nest densities throughout an area, and maintaining higher populations. Likewise, P. megacephala would be expected to have superior interference abilities displayed as direct aggression, both at the individual and colony level. Evidence to support all of these characteristics can be inferred from the literature, but only the mechanism of higher population levels has been formally quantified (e.g., Majer and de Kock 1992; Hoffmann et al. 1999; Vanderwoude et al. 2000).
An overall reduction in macro-invertebrate abundance was found relative to time since invasion (I1–I4) by P. megacephala, and this impact was almost uniform among different invertebrate groups such that no ordinal groups had been entirely lost from infested areas, with the exception of Orthoptera in the oldest infested site. However, there were also some clear positive associations with P. megacephala, particularly isopods, cockroaches, and bugs. Isopods and cockroaches are well known to persist in the presence of invasive ants most likely due to their armoured integuments (Haines and Haines 1978; Rao et al. 1989; Lester and Tavite 2004) and many bugs readily form symbiotic relationships with this ant (Holway et al. 2002 and references therein). It should be noted that it is likely that more pronounced effects on the macro-invertebrates would have been detected with analysis at Family or species level.
These clear impacts found by pitfall trap data from 2005 concur with the findings from foliage beats and litter samples in 1996, as well as with the general understanding of the ecological impacts of invasive ants (e.g., Porter and Savignano 1990; Williams 1994 and references therein), particularly on invertebrate faunas that have evolved in the absence of ants (Perkins 1913; Gillespie and Reimer 1993). But despite its global distribution and well-noted ecological impacts, only one other study has quantified P. megacephala impacts on non-ant invertebrates (Collembola) within a relatively undisturbed environment, and it found the ant had no overall effect (Heterick 1997). A similar result was also found by Holway (1998) studying the impacts of the ecologically similar exotic ant L. humile within intact riparian woodlands in California, despite a 4- to 10-fold increase in ant density and an almost complete extirpation of the native epigaeic ant fauna. This led to the conclusion that the invasive ant and the native ants they displace interact with the ground-dwelling arthropods of these habitats in a similar manner. Both the studies of Heterick (1997) and Holway (1998) investigated impacts on systems containing native ant faunas considered to be competitive, whereas here, and within Australia’s rainforests in general, the native epigaeic ant faunas are poorly competitive (Andersen 1995). This supports the idea that the magnitude of the impacts of exotic ants on native non-ant invertebrates is likely to be an artefact of the evolutionary coexistence of the invertebrates with aggressive native ant species (Medeiros et al. 1986; Gillespie and Reimer 1993), a notion that warrants closer attention globally.
There is no doubt that P. megacephala is a serious ecological threat throughout northern Australia, just as it is considered to be globally (Williams 1994). Its populations display no sign of stabilisation or restriction in range after more than at least three decades at Howard Springs and its ecological impacts remain severe. This ant is relatively easy to eradicate from isolated areas (Hoffmann and O’Connor 2001), and we have shown that there is an urgent need for the eradication of this ant at Howard Springs. These results should also give cause for serious concern where this ant is known to occur within other conservation areas world-wide, and we believe that a larger management focus is required on this species globally.
We thank research students Stephanie Petitcunot and Juliette Payet for their assistance, and the Northern Territory Parks and Wildlife staff at Howard Springs Nature Park for access to the reserve and support for the project. Comments by Alan Andersen, Lori Lach, Kirsti Abbott, Paul Krushelnycky and an anonymous reviewer improved the manuscript. Additional funding was provided to CLP from the Trapnell fellowship. This research was conducted under Northern Territory Parks and Wildlife permit 25506.